There is one other factor to consider.

When following the instructions to install your own NTE5 after a pre-existing LJ2 (the diagram of which was provided by Clarity UK and used by Kitz, with Clarity's consent), the circuit will then have two resistive-capacitive shunts connected across the pair. One shunt being in the original LJ2 and the second shunt in the NTE5.

That was my thinking when I posted earlier.

What would be the effect of having that on the line?

So you do actually have two masters on the circuit, which is not a good idea. Whether it can cause this problem, or even be having an overall deleterious effect, I'm not qualified to say.

At a guess, wiring from the back of the LJ2/1A has created a star-wired installation. It's front socket on one spur and the NTE5 plus its extensions on the other.

Or, now I look at your linked diagram, do you mean you used as shown the extension terminals 2 and 5 on the LJ2/1A - not the real back of it where the incoming line is connected? That wouldn't be star wiring so far as the LJ2/1A is concerned, but it may be as far as the NTE5A is concerned.

At which point my knowledge of the consequences runs out.

What, if anything, happens if you try disconnecting the NTE5 from the back of the LJ2/1A, then plugging the modem into the LJ2/1A's socket?

If it's star wiring issues, the shift to the LJ2/1A as only socket, with you in the front of it, should fix it.

Unfortunatly the original master socket is behind a large cupboard, and I am struggling to find the photos I took back when the NTE5 was fitted.

If my memory is correct, the incoming line and the Cat5e are both connected to the same set of terminals on the back of the original master socket.

Basically like this image I just found via Google Images, minus the ring wire in 3.

http://i126.photobucket.com/albums/p89/bomp_bucket/m... (Note: This is NOT my socket)

I was forgetting the LJ2/1A doesn't have separate A/B incoming line connectors, but that pic looks like one wired as an extension, with split pairs!

It's a good job it isn't yours.

How yours is wired matters. I assume you haven't used split pairs in your CAT5, anywhere.

What do you mean by split pairs?

Slightly higher attenuation, possibly in the voice band frequencies, would be my best guess.

The R/C shunt consists of a 1.8 micro Farad capacitor and a 470 kilo Ohm resistor connected in series. The simplest solution would be to remove one of the shunts . . . the one within the NTE5 that you installed.

The purpose of the that R/C shunt is to provide an out-of-service termination for the circuit when no equipment is connected across the pair. Such shunts came into being when the current style, parallel connection, plug and socket arrangement was adopted in the early 1980s.

In cat5 wires are twisted together e.g. blue with a thin white stripe, and white with a thin blue strip, those form a pair.

A split pair is when you wire the orange with thin white to one terminal, and the blue with thin white to another, ie. use a single wire from two pairs.

The split means that the noise rejection that the twists in the cable produce are lost, so cable does not perform as well.

Ah got it. No split pairs on my wiring, everything is using the blue and blue/white wires (pair 1).

As an IT Professional I understand the importance of twisted pairs.